by
Teesha Crystal Luehr
Bachelor of Science (Honours), University of British Columbia, 2015
A Thesis Submitted in Partial Fulfillment of the Requirements for the Degree of
MASTER OF SCIENCE
in the Department of Biochemistry & Microbiology
ã Teesha Crystal Luehr, 2017 University of Victoria
All rights reserved. This thesis may not be reproduced in whole or in part, by photocopy or other means, without the permission of the author.
Multiplexed Matrix-Assisted Laser Desorption/Ionization-Mass Spectrometry Imaging (MALDI-MSI) Biomarker Discovery
by
Teesha Crystal Luehr
Bachelor of Science Honours in Chemistry, University of British Columbia, 2015
Supervisory Committee
Dr. Christoph H. Borchers, Department of Biochemistry and Microbiology Supervisor
Dr. Caren C. Helbing, Department of Biochemistry and Microbiology Departmental Member
Dr. Ben F. Koop, Department of Biology Outside Member
Supervisory Committee
Dr. Christoph H. Borchers, Department of Biochemistry and Microbiology Supervisor
Dr. Caren C. Helbing, Department of Biochemistry and Microbiology Departmental Member
Dr. Ben F. Koop, Department of Biology Outside Member
The work presented herein is a method optimization for biomolecule detection and identification using Matrix-Assisted Laser Desorption/Ionization-Mass Spectrometry Imaging (MALDI-MSI). MALDI-MSI is a unique form of mass spectrometry that is highly multiplexed; it can simultaneously retain location information of the mass of multiple ions, allowing for correlation of morphology or pathology to reconstructed ion heat maps. There were three main objectives for the research - 1) A method optimization of sample preparation techniques for bottom-up proteomic MALDI-MSI was performed. This included the optimization of tissue wash steps, trypsin digestion incubation times, and matrix deposition techniques. The results included identifying the appropriate pH for the wash steps to optimize trypsin digestion, an overnight trypsin incubation to allow for complete digestion, and the inclusion of MCAEF – Matrix Coating Assisted by an Electric Field – during matrix coating for enhanced spectra. 2) An unbiased statistical data processing workflow for simultaneous processing of multiple datasets was
performed. This was done using a thyroid hormone treated tadpole dataset to gain insight into the metabolism of anuran metamorphosis. Results found included a finalized data processing workflow that detected 5000 metabolite features from five organs were detected in pre-metamorphic tadpoles. Of these detected metabolites, 136 were significantly affected upon exposure to thyroid hormone and 64 metabolites were putatively identified. 3) A sample preparation technique for metabolomic analysis of formalin-fixed paraffin embedded (FFPE) colorectal liver metastasis samples was performed. Results included the importance of using a high mass resolution mass spectrometer while emphasizing more appropriate use of fresh-frozen tissue sections for metabolomic analysis.
Supervisory Committee ... ii
Abstract ... iii
Table of Contents ... iv
List of Tables ... v
List of Figures ... vi
List of Tables in Appendix ... ix
Dedication ... x
List of Abbreviations ... xi
Thesis Format and Manuscript Claims ... xiv
Chapter 1: Introduction ... 1
The Need for Biomarkers ... 1
Mass Spectrometry ... 4
Matrix Assisted Laser Desorption/Ionization ... 5
MALDI-Mass Spectrometry Imaging ... 6
Data Processing ... 14
Bottom-Up Proteomics & Metabolomics in MALDI-MSI ... 19
Challenges in the Field of MALDI-MSI ... 20
Thesis Objectives ... 20
Chapter 2: Method Optimization of Bottom-Up Proteomic Analysis of Prostate Cancer by MALDI-MSI for Biomarker Discovery ... 22
Introduction ... 22
Methods... 26
Results & Discussion ... 38
Conclusions & Future Directions ... 48
Chapter 3: Metabolomic insights into the effects of thyroid hormone on Rana catesbeiana metamorphosis using whole-body Matrix Assisted Laser Desorption/Ionization-Mass Spectrometry Imaging (MALDI-MSI) ... 50
Introduction ... 50
Methods... 54
Results & Discussion ... 57
Conclusions & Future Directions ... 74
Chapter 4: Method Optimization of Metabolomic Analysis of Formalin Fixed Paraffin Embedded Colorectal Liver Metastasis by MALDI-MSI for Biomarker Discovery ... 76
Introduction ... 76
Methods... 79
Results & Discussion ... 82
Conclusions & Future Directions ... 87
Chapter 5: Conclusions & Future Directions ... 89
Bibliography ... 90
Table 1 – A summary of recently published novel matrices for MALDI-MSI and their corresponding biological applications adapted from Baker et al. 2016. ... 10 Table 2 – A summary of bioinformatic tools relevant to MALDI-MSI for use in data conversion, visualization, basic spectral process, and statistical analysis modified from Baker et al. 2016. Abbreviations: F indicates a free software; P indicates a paid for
software; B indicates a paid for MATLAB package; G indicates a downloadable graphical
user interface software; R indicates a R Statistical package; M indicates a MATLAB package; W indicates an online web server ... 17 Table 3 – A summary of the number of metabolite mass features detected in each tissue type in all tadpoles in positive mode. *Total number of mass features detected in a minimum of 3 tadpoles in each of the control and treatment groups. **Significantly different mass features based on Mann-Whitney U post-hoc analysis with a p-value < 0.05... 64
Figure 1 – Image depicting the relationship between genomics, proteomics, and metabolomics and their increasing environmental influence.3 Image reprinted with
permission. ... 2 Figure 2 – A schematic diagram showing the relationship between the genome,
transcriptome, proteome, and metabolome. The metabolome which is comprised of sugars, nucleotides, amino acids, and lipids, make up a phenotype of an organism.
Retrieved from https://en.wikipedia.org/wiki/File:Metabolomics_schema.png ... 3 Figure 3 - MALDI-MSI sample preparation workflow including tissue sectioning, matrix coating, laser irradiation, spectral acquisition, and ion heat map image reconstruction. ... 8 Figure 4 – A diagram of the proposed mechanism of the MCAEF technique developed and used in the Borchers laboratory. Two electrically conductive glass slides facing each other have an applied voltage (600V/m) creating a uniform electric field. This field causes a micro-extraction of charged particles to the surface of the tissue to crystalize with the matrix. Image adapted from Baker et al. 2016.36 ... 13 Figure 5 - Reconstructed ion heat maps of proteins shown to have differential expression in the cancerous and non-cancerous regions of a human prostate cancer tissue. Proteins were detected with sinapinic acid as the matrix in positive mode58. Image also adapted from Baker et al. 2016.36 Images are reprinted with permission. ... 23 Figure 6 – Reconstructed ion heat maps of metabolites (phospholipids and neutral lipids) shown to be of interest in a human prostate cancer tissues. Metabolites were detected with quercetin as the matrix in both positive and negative ionization mode39. Images are reprinted with permission. ... 24 Figure 7 – Workflow used for accurate mass matching between MALDI-MSI and LC-MS/MS data for peptide identification and protein assignment ... 37 Figure 8 – (a) Average peak intensity and (b) average peak count of the MALDI-MSI wash optimization experiment on prostate cancer tissue. Detail for each wash can be found in the Methods section. Not all significant comparisons shown on graph. Detailed p-values can be found in Supplementary Table 2.* p < 0.05, ** p < 0.01, and *** p < 0.001... 38 Figure 9 – (a) Average peak intensity and (b) average peak count of the MALDI-MSI trypsin digestion incubation time experiment on prostate cancer tissue. Details for each incubation time can be found in the Methods section. Detailed p-values can be found in Supplementary Table 2.* p < 0.05, ** p < 0.01, and *** p < 0.001 ... 40 Figure 10 – The scores plot from a Principle Component Analysis test for the MALDI-MSI bottom-up proteomic prostate cancer optimization experiment. Data are with MCAEF and without MCAEF during matrix deposition. Details on the MCAEF set up can be found in the Methods section. ... 42 Figure 11 – The scores plot from a Partial Least Squares-Discriminant Analysis test for the MALDI-MSI bottom-up proteomic prostate cancer optimization experiment. Data are with MCAEF and without MCAEF during matrix deposition. Details on the MCAEF set up can be found in the Methods section. ... 42 Figure 12 – (a) Average peak intensity and (b) average peak count of the MALDI-MSI MCAEF optimization experiment on prostate cancer tissue. Details for the MCAEF set
Supplementary Table 2.* p < 0.05, ** p < 0.01, and *** p < 0.001 ... 44 Figure 13 – Reconstructed ion heat maps from the bottom-up proteomic MALDI-MSI prostate cancer experiment. Peptide mass features shown differentiate between the cancerous and non-cancerous regions. The size and intensity scale, along with an H&E image can be found in the bottom right hand corner. Abbreviations: CD – cold-inducible RNA binding protein; HB – hemoglobin; P1RS – phosphatase 1 regulatory subunit; PSA – prostate specific antigen; T S/R – thiosulfate sulfurtransferase/rhodanese-like domain containing protein ... 46 Figure 14 – Representative bottom-up proteomic mass spectra acquired MALDI-MSI from (top) cancerous and (bottom) non-cancerous regions of the prostate cancer tissue. 47 Figure 15 – (Left) A MALDI-MSI workflow depicting tissue sectioning onto a glass slide, homogenous matrix coating using the MCAEF technique, rastered laser irradiation across the tissue, and ion heat map image reconstruction based on the acquired mass spectra (Right) A diagram of the MCAEF technique during matrix coating to enhance both the total number of features detected and their corresponding intensities for a single experiment. ... 53 Figure 16 - A representative image of spectra processing performed by the Cardinal package on the tadpole imaging data. Spectra shown are average spectra across the entire tissue. ... 58 Figure 17 – A representative image of Cardinal reconstructed ion heat maps during spectral processing performed by the Cardinal package on the tadpole imaging data. .... 59 Figure 18 – Cardinal spatial segmentation images using different values for the r (radius) parameter and k (number of segments) parameter during spatial segmentation
optimization. ... 61 Figure 19 – Representative exports from the Cardinal package including (top image) the total segmentation; (second image) an individual segment, for this example it is the brain; (third image) the mean spectrum from the individual segment from the shrunken spatial segmentation; and (bottom image) the t-statistic value for each mass feature relative to the individual segment ... 63 Figure 20 – Representative reconstructed ion heat maps of metabolites that localize to a single tissue of interest. Ion heat maps were chosen from any of the twelve tadpoles as a representation of the localization of the metabolite mass feature. ... 65 Figure 21 – Representative boxplot graphs of significant (MWU, p < 0.05) metabolite mass features from: A) brain, B) eye, C) liver, D) notochord, and E) tail muscle tissues. Feature characteristics including m/z, putative identifications are indicated at the top of each graph. The thick bar represents the median and the whiskers represent the median absolute deviation (MAD). ... 66 Figure 22 – Representative reconstructed ion heat maps of metabolites from each tissue region of interest (brain, eye, liver, notochord, and tail muscle). One ion heat map was chosen from the set of control tadpoles and one ion heat maps was chosen from the set of thyroid hormone treated tadpoles. Each metabolite mass feature chosen was found to be significantly increased or decreased after thyroid hormone treatment (p < 0.05).
Visualization of the mass features shows the localization of each metabolite to a specific tissue region and its increase or decrease in the treatment tadpoles. ... 67
maps of metabolites of special biological interest. Feature characteristics including m/z, putative identifications are indicated at the top of each graph. The thick bar represents the median and the whiskers represent the median absolute deviation. One ion heat map was chosen from the set of control tadpoles and one ion heat maps was chosen from the set of thyroid hormone treated tadpoles. Each metabolite mass feature chosen was significantly increased or decresed after thyroid hormone treatment (p < 0.05). ... 69 Figure 24 – MALDI-MSI metabolomic images of (A) a colorectal liver metastasis sample analyzed with quercetin in positive ionization mode and (B) a colorectal liver metastasis sample analyzed with quercetin in negative ionization mode. Images include the optical image, an H&E stained image, total segmentation image, the individual cancerous segment image, and a representative reconstructed ion heat map of an influential
metabolite mass feature. ... 84 Figure 25 – Representative spectra from MALDI-MSI metabolomic dataset of (A) a colorectal liver metastasis sample analyzed with quercetin in positive ionization mode and (B) a colorectal liver metastasis sample analyzed with quercetin in negative
Supplementary Table 1 – Summary of literature search for method protocol steps for preparing a tissue for bottom-up proteomic MALDI-MSI analysis. ... 108 Supplementary Table 2 – MALDI-MSI bottom-up proteomic analysis Mann-Whitney U test p-values. ... 113 Supplementary Table 3 – Summary of recently published proteomic MALDI-MSI
experiments on FFPE tissues. RT – room temperature; ACN – acetonitrile; secs –
seconds; min – minutes; hr – hours; TFA – trifluoroacetic acid ... 114 Supplementary Table 4 – Continuation of Supplementary Table 3 – Summary of recently published proteomic MALDI-MSI experiments on FFPE tissues. ... 117 Supplementary Table 5 – Continuation of Supplementary Table 3 and Supplementary Table 4 - Summary of recently published proteomic MALDI-MSI experiments on FFPE tissues. ... 120 Supplementary Table 6 – Summary of all bottom-up proteomic m/z values found in prostate cancer by MALD-MSI accurate mass matched to LC-MS/MS data for peptide identification and corresponding protein assignments for fresh frozen prostate cancer . 123 Supplementary Table 7 – Complete list of metabolite mass features found from MALDI-MSI in a minimum of three control and three thyroid hormone treated tadpoles that showed significant (p<0.05) difference between control and treatment tadpoles with corresponding putative identifications, if found. The medians ± median absolute
I would like to take this opportunity to thank three people for their continued dedication and support during this research: 1) I would first like to say a special thank-you to my husband, Stephen Luehr. You inspire me to reach my greatest potential in life. Together we will reach the moon, and if not, we’ll land somewhere in the stars. 2) Secondly, I would like to thank my 10-year old self. She was an amazingly dedicated young lady who always said she wanted to grow up to be wearing a lab coat and goggles. Here I am today, writing my Masters thesis and about to undertake a PhD, all in
biochemistry. The momentum she began brewing has helped power me through the trials and tribulations of this graduate degree. 3) Last but never the least, I would like to thank the graduate students of BCMB. You have stood by my side and helped me grow into not only a stronger researcher, but a stronger person. I am grateful for the lessons learned, the laughs shared, and the drinks consumed! And of course, a quote that has helped me get through this thing we call life: If at first you don’t succeed, laugh until you do.
2-MBT - 2-mercaptobenzothiazol 3-HC - 3-hydroxycoumarin 9-AA - 9- aminoacridine ACN - acetonitrile
ATT - 6-aza-2-thiothymine BRCA - breast cancer 1 gene CA125 - cancer 125 protein
CD - cold-inducible RNA binding protein CEA - carcinoembryonic antigen
CHCA - α-cyano-4-hydroxycinnamic acid CID - collision induced dissociation DAN - 1,5-diaminonapthalene DHAP - 2,5-dihydroxyacetophenone DHB - 2,5-dihydroxybenzoic
DMAN - 1,8-bis(dimethyl-amino) naphthalene DNA - deoxyribonucleic acid
DT - dithranol DTT - dithiothreitol
ESI - Electrospray Ionization FA - formic acid
FFPE - formalin-fixed and paraffin embedded FTICR - Fourier transform ion cyclotron resonance
H&E - hematoxylin & eosin HB - hemoglobin
HCD - higher-energy collisional dissociation HLA - Human Leukocyte Antigen
HPLC - high performance liquid chromatography hr - hours
IAA - iodoacetamide ITO - indium-tin oxide
LC MS/MS - Liquid Chromatography-Tandem Mass Spectrometry LC/MS - liquid chromatography/mass spectrometry
LC-MS/MS - liquid chromatography tandem mass spectrometry MAD - median absolute deviation
MALDI - matrix-assisted laser desorption/ionization
MALDI-MSI - Matrix-Assisted Laser Desorption/Ionization-Mass Spectrometry Imaging MCAEF - Matrix Coating Assisted by an Electric Field
min - minutes
MWCNT - Multi-Walled Carbon Nanotube NaOH - sodium hydroxide
NH4HCO3 - ammonium bicarbonate
NH4OH - ammonium hydroxide
P1RS - phosphatase 1 regulatory subunit PC - phosphatidylcholine
PE - phosphatidylethanolamine PG - phosphatidylglycerol PI - phosphatidylinositol
PLS-DA - partial least squares-discriminant analysis PS - phosphatidylserine
PSA - prostate specific antigen RNA - ribonucleic acid
SA - sinapinic acid secs - seconds SM - sphingomyelin
SPE - solid phase extraction
T S/R - thiosulfate sulfurtransferase/rhodanese-like domain containing protein T4 - thyroxine
TFA - trifluoroacetic acid TH - thyroid hormone TOF - time-of-flight
The chapters in this thesis are presented in the format of a manuscript. The first chapter provides a general background and context for the thesis and introduces the rationale behind the thesis objectives. Chapters 2, 3, and 4 are written in a manuscript style and contain an Introduction, Materials & Methods, Results & Discussion, and Conclusions & Future Directions. The final chapter summarizes the data chapters.
Chapter One – Subheading MALDI-MSI: Baker, T.C., Han, J., Borchers, C.H., 2016. Recent advancements in matrix-assisted laser desorption/ionization mass spectrometry imaging. Current Opinion in Biotechnology 43, 62–69. doi:10.1016/j.copbio.2016.09.003 Baker, T.C. performed the literature review wrote the initial manuscript. Han, J. assisted in editing and finalizing the manuscript.
Chapter Three: Luehr, T.C., Koide, E.M., Wang, X., Han, J., Borchers, C.H., Helbing, C.C., 2017. Metabolomic insights into the effects of thyroid hormone on Rana
catesbeiana metamorphosis using whole-body Matrix Assisted Laser
Desorption/Ionization-Mass Spectrometry Imaging (MALDI-MSI). General and
Comparative Endocrinology. Submitted to a special edition issue. Helbing, C.C. designed the exposures. Wang, X. acquired the data under supervision of Han J. Luehr, T.C. performed all data processing and statistical analysis. Luehr, T.C. created all figures and tables for the publication. Luehr, T.C. wrote the manuscript with assistance from Koide, E.M. and Helbing, C.C.
Chapter 1: Introduction
The Need for Biomarkers
Cancer has very few clinical biological markers for aiding in detection and diagnosis of the disease. These biological markers, can be in the form of genetic, transcriptomic, proteomic, and metabolomic biomarkers. Optimally, a biomarker or panel of biomarkers can distinguish minute differences between related diseases that may be difficult to diagnose by a medical practitioner.
The relationship between genetics, transcripts, proteins, and metabolites is diverse. There is an intimate relationship between genes and their eventually affected metabolites. It has been shown that a change in a single base pair in a gene can result in a 10,000 fold change in a metabolite’s concentration in the human body.1–3 Figure 1 shows a schematic
diagram of the relationship between genomics, proteomics, and metabolomics. This figure depicts metabolites at the top of the pyramid, indicating its strong influence in our environment and within the human body. Metabolomics are extremely valuable in clinical diagnostics of diseases other than cancer. Examples that include the involvement of metabolites or metabolism include: diagnostic clinical assays (95%)4, known drugs
(89%)5, drugs derived from pre-existing metabolites (50%)6, and genetic disorders (30%)7.
Figure 1 – Image depicting the relationship between genomics, proteomics, and metabolomics and their increasing environmental influence.3 Image reprinted with permission.
The metabolomic profile is extremely sensitive to changes at the genomic and proteomic levels, resulting in the metabolome acting as a measurable phenotype within an organism (Figure 2). Understanding how the metabolome can give insights into cancer phenotype is a topic of recent research.8–14 A very large review on lipid signalling,
specifically phosphoinositides (PI), has been performed in the last 15 years on the universal impact of lipid signalling in eukaryotes.15
Figure 2 – A schematic diagram showing the relationship between the genome,
transcriptome, proteome, and metabolome. The metabolome which is comprised of sugars, nucleotides, amino acids, and lipids, make up a phenotype of an organism. Retrieved from https://en.wikipedia.org/wiki/File:Metabolomics_schema.png
Genetic biomarkers are often used to determine the level of risk a patient has for developing a disease. For example, diabetes has multiple genes in the human leukocyte antigen (HLA) and a variable number of tandem repeats in the insulin gene have both been correlated with type 1 diabetes.16 Breast cancer has mutations to the breast cancer 1
(BRCA1) and BRCA2 genes have served as risk factors and guides for targeted therapy.17 Proteomic biomarkers, often detected in a patient’s serum, are used for early diagnosis of diseases. Cancer 125 (CA125 - a mucin glycoprotein) concentration is monitored in
serum to determine the risk of developing ovarian cancer and aids in determining effective therapy.18 Metabolomic biomarkers are the key to connecting genetics to the visible phenotype. Recently, a set of metabolomic biomarkers has been proposed for predicting the risk of developing pancreatic cancer.19 For example, prostaglandin E2 metabolites have been correlated with pancreatic cancer risk.19 These metabolites were
discovered as potential biomarkers due to their presence in the cyclooxygenase 2 pathway, which is known to be upregulated in pancreatic cancer.19
The biomarkers discussed above are examples of molecules that can be used in screening, diagnosis, prognosis, and targeted therapy choices of various diseases. However, the limited number of cancer biomarkers have poor specificity in
distinguishing between sub-types of cancers. Cancer biomarkers require further study for development of effective clinical tests.
Mass Spectrometry
Mass spectrometry (MS) is an extremely dynamic tool that has the ability to detect nearly any type of compound in a biological sample.20 MS can determine the exact masses of individual compound which can lead to an identification of the biomolecule.21 An advantage of using mass spectrometry over other histological methods is that it can concurrently analyze a wide range of masses at the same time, creating a multiplexed analysis within a single tissue.22 MS as a whole is a valuable technique for use in
biomarker discovery and other untargeted methods aiming to characterize unknown compounds in biological samples.23–28
In general, mass spectrometers are comprised of three main components: an ionization source to create ions from the introduced sample, an analyzer to separate ions by certain mass characteristics and mass-to-charge ratios, and a detector for recording the
abundancies of each mass-to-charge ratio.29 The two most common ionization sources for proteomic MS analysis are electrospray ionization (ESI) and matrix-assisted laser
desorption/ionization (MALDI).29 The most essential component of a mass spectrometer is the mass analyzer.29 There are four basic types of mass analyzers: ion trap,
time-of-flight (TOF), quadrupole, and Fourier transform ion cyclotron resonance (FTICR) analyzers.
Matrix Assisted Laser Desorption/Ionization
MALDI is a type of mass spectrometry ionization source. MALDI was originally introduced in 1988 as an ionization source as a solution to ionize large macromolecules, such as intact proteins.30 It is useful for proteins and other large compounds because it is considered a soft ionization source; the desorption and ionization process result in singly charged ions that typically aren’t fragmented from their original state. MALDI is unique to many other ionization techniques in that the sample is introduced to the instrument in a solid state.
As described in the name, this ionization source utilizes a matrix for aiding in both desorption and ionization of the molecules in the biological sample. A matrix is a
compound that contains a chromophore that readily absorbs light at the wavelength of the laser.31 The absorption of radiation causes a vibrational excitation of the matrix, leading to desorption of the matrix and neighbouring non-matrix compounds into the gas phase. Simultaneously, the excitation causes ions to be formed. Many different matrices have been experimented with, each with their own distinct benefits and negative attributes.
MALDI-Mass Spectrometry Imaging
Creation
Ten years after the creation of MALDI as an ionization source, Caprioli and his laboratory group applied this technique directly onto tissue sections for simultaneous in-situ detection of biomolecules, known as MALDI mass spectrometry imaging (MALDI-MSI).32 This application spatially maps the concentration of an analyte by retaining
location information on the tissue. A laser is rastered across a tissue section acquiring a full mass spectrum at each laser pixel (Figure 3). As a result, ion heat maps can be reconstructed for each tissue, displaying the relative concentration of an ion across the tissue section. The relative concentration information can be correlated to biology, morphology, histology, or pathology for a greater understanding of the biological context of each ion detected.
Methodology
A visual representation of a MALDI-MSI workflow can be found in Figure 3. In short, the general method applied during a MALDI-MSI experiment is as follows: a biological sample is thinly sliced (5-50 µm) with a microtome or cryostat, the tissue section is placed onto an electrically conductive surface such as a metal MALDI plate or a indium-tin oxide (ITO) coated glass slide. Depending on the type of compounds to be detected there may be some form of on-tissue sample preparation. For example, for detection of tryptic peptides an on-tissue digestion must be performed33, or for detection of peptides or proteins a small molecule wash must be performed34. Next, the tissue is thinly coated with a matrix using a homogenous matrix sprayer or generator. Finally, the tissue is placed into the mass spectrometer for laser irradiation and detection. Each of these steps requires a wide range of optimization based on the tissue type, the class of compounds of interest, and the instrument being used.
Figure 3 - MALDI-MSI sample preparation workflow including tissue sectioning, matrix coating, laser irradiation, spectral acquisition, and ion heat map image reconstruction.
Matrices
A matrix is used in MALDI to aid in increased desorption of analytes of interest from the conductive surface.35 Matrices are commonly a UV-absorbing aromatic acids.35 The compounds absorb the laser pulse resulting in desorption and energy transfer to the analyte for ionization.35 A list of older, yet still commonly used matrices includes: 2,5-dihydroxybenzoic (DHB), α-cyano-4-hydroxycinnamic acid (CHCA), sinapinic acid (SA), 2-mercaptobenzothiazol (2-MBT), 9- aminoacridine (9-AA), and
2,5-dihydroxyacetophenone (DHAP). One of the most well studies areas of method
optimization in MSI is determination of which matrix to use for each MALDI-MSI experiment. A recent review in novel matrices was performed.36 A summary of their
application and uses can be found in Table 1. All matrices discussed in this table are for biological use in detecting and quantifying small molecules. This is a result of two major factors: 1) recent MALDI-MSI research has led to a large increased focus on small molecules including lipids and other metabolites, and as a result there is an increased amount of research on useful matrices and 2) older matrices that were optimal for detection of large molecules often had very complex background signals in the low molecular weight region (100-1000 m/z), which created difficulty in detecting small molecules occurring in the same region of the mass spectrum. One matrix of particular note is Quercetin.37 This matrix was shown to have low volatility under vacuum, low matrix-related ions, and low threshold for laser desorption and ionization.37 After initial
publication, quercetin has been used in other experiments.38–40 The final results of the experiment found that over 500 lipids were detected in positive and negative ionization mode in porcine adrenal glands, which was the largest number of lipids analyzed in a single MALDI-MSI study at the time of publication.40
Table 1 – A summary of recently published novel matrices for MALDI-MSI and their corresponding biological applications adapted from Baker et al. 2016.
(Ionization mode)-Matrix Application Instrument Ref
(+/-)1,5-diaminonapthalene (DAN) Higher sensitivity of ions in positive mode and rich spectra in negative mode for lipid analysis.
TOF/TOF 41
(+)1,7-bis-(4-hydroxy-3-methoxy-phenyl)-hepta-1,6-diene-3,5-dione (Curcumin)
Versatile and multipurpose analysis of pharmaceutical drugs, lipids, peptides, and proteins by promoting ionization
QTOF and HDMS 42
(-)1,8-bis(dimethyl-amino) naphthalene (DMAN) A clean background spectra for analysis of metabolites in negative mode.
TOF/TOF 43
(-)1,8-bis(dimethylamino)naphthalene(DMAN)/9-aminoacridine (9-AA)
Reduced chemical noise, and no matrix-clusters for lipid analysis
TOF 44
(+)3-hydroxycoumarin (3-HC)/6-aza-2-thiothymine (ATT) Low amount of background signals for analysis of drugs and single amino acids
TOF 45
(-)4-phenyl-α-cyanocinnamic acid amide Small number of background peaks with matrix suppression for analysis of various lipid classes.
TOF/TOF 46
(+) Alternating Multilayer of (-) Graphene oxide (GO) and (+) Multi-Walled Carbon Nanotube (MWCNT)
Electrostatic charge between layers for removal of
interference and contamination for analysis of small molecules
(+) Dithranol (DT) Higher detection of endogenous lipids in positive mode over commonly used matrices.
FTICR 49,50
Quercetin Lower number of matrix-related
ions with a higher detection of lipids.
FTICR 37
(+) Two-Dimensional Graphene Minimal background
interference from elemental carbon.
Matrix Deposition
Another area of MALDI-MSI examined in the review referenced above was novel matrix deposition techniques.36 The overall quality of matrix deposition highly affects the spatial resolution in a MALDI-MSI experiment; matrix deposition needs to be uniform and create relatively small particle sizes for the laser irradiation to be effective.
Commonly used matrix deposition techniques include matrix solution spray coating, matrix sublimation, and matrix pre-coating. Matrix spray coating involves a manual or automatic electronic or heat assisted spray generator to uniformly deposit the matrix on top of the tissue. This technique typically creates small, uniform particle sizes (10-15 µm).52 Matrix sublimation has gained in use in laboratories for its ability to create low-µm particle sizes in a solvent-free manner.53,54 Sublimation is compatible with all of the previously mentioned common matrices, but requires in-lab optimization of each matrix to achieve optimal performance.55 The final matrix deposition technique is rather unique and different from typical matrix deposition techniques. The matrix is applied prior to the addition of a tissue section.56,57 Sublimation reduced matrix deposition time and useful for both high molecular weight analytes such as proteins using SA and for low molecular weight analytes such as lipids and peptides using DHB.56
Matrix Coating Assisted by an Electric Field
A novel matrix deposition technique was developed in the Borchers laboratory in 2015: Matrix Coating Assisted by an Electric Field (MCAEF).38 A diagram of the proposed
mechanism can be found in Figure 4. This technique shows a significant increase in both the total number of analytes detected in a MALDI-MSI experiment and the signal to
noise ratio of those analytes. This is highly beneficial for untargeted experiments that are aiming to detect large number of analytes. The original proof-of-concept experiment was performed on rat brain and detected and identified 648 lipids in both positive and
negative mode as compared to only 344 lipids in both positive and negative mode without MCAEF. In addition, 232 proteins were detected in positive mode as compared to only 199 proteins in positive mode without MCAEF. This technique has been further applied to human prostate cancer for use in intact protein biomarker discovery. A total of 17 proteins have been detected and determined to have significantly different distribution patterns between the cancerous and non-cancerous tissue regions.58
Figure 4 – A diagram of the proposed mechanism of the MCAEF technique developed and used in the Borchers laboratory. Two electrically conductive glass slides facing each other have an applied voltage (600V/m) creating a uniform electric field. This field causes a micro-extraction of charged particles to the surface of the tissue to crystalize with the matrix. Image adapted from Baker et al. 2016.36
Mass Analyzers
There are two commonly used mass analyzers for MALDI-MSI: TOF and FTICR. An FTICR mass analyzer is an ion trap style mass analyzer that applies a uniform electric field that matches the cyclotron frequency of the ions in the magnetic field.59 The
uniform electric field excites the ions into a larger orbit that will pass by a detector.59 The
frequencies measured by the detector are converted to m/z using a Fourier transform.59 A TOF mass analyzer accelerates ions using an electric field.59 Once ions are at the same potential energy, they advance down a tube where smaller ions reach the detector before larger ones.60 As a result of their physical design, FTICR analyzers have a higher
resolving power for low molecular weight ions, whereas TOF analyzers excel in detecting high molecular weight ions. A recent research publication directly compared the applicability of the two detectors on the same metabolite MALDI-MSI experiment.61 The conclusion remained the same, instruments with a FTICR detector are far better at resolving and detecting low molecular weight ions, with up to 5 times more low m/z features detected by the FTICR.61 Both FTICR62 and TOF63 have shown to be highly
useful tools for biomarker detection using for cancer using MALDI-MSI.
Data Processing
The importance of informatics in research is often underestimated. The need for data processing and interpretation is universal across different areas of research, countries, and languages. The ability to extract meaning from data would be stunted without informatics as acquired data would be restricted to the simplest forms or be left unprocessed.
Informatics is the science and engineering of information systems for processing, analyzing, and implementing data. Bioinformatics is specifically focused on software tools for the processing and understanding of biological data. Bioinformatics tools use computer science, mathematics, statistics, and the biological sciences. Computer scientists, mathematicians, and statisticians create and optimize tools for biological scientists. An open line of communication is needed to develop software for specific purposes. Research labs are becoming vastly interdisciplinary providing the advantage of direct communication between the makers and the users of the software. Informatics tools and statistics have languages of their own and are not divided by continents, but rather by fields of research. Simple statistics, such as standard deviation of the mean, are common among all scientists, but certain statistics used in genomics may not be applicable to microbiologists or bioengineers. For this reason, bioinformatic tools are often limited to a specific niche. However, this becomes beneficial because each piece of software can be highly specialized and optimized for selective data input.
MALDI-MSI produces large data sets that require extensive bioinformatics analysis. Manually searching and mining the data would take an extensive amount of time and would be plagued with user bias. Basic analysis and visualization of heat maps is the first step to processing MALDI-MSI data. Next is reducing the number of heat maps for a dataset to those that are statistically relevant to the experiment. The most commonly mentioned analysis and visualization software are provided by the instrument vendors: FlexImaging (Bruker), ImageQuest (Thermo Fisher Scientific), MSImageView
(Novartis), and Quantinetix (Imabiotech). Unfortunately, few statistical analysis tools are available. SCILS Lab from Bruker and SCILS is an add on software that works
seamlessly with FlexImaging to provide multivariate analyses of MSI data. However, both FlexImaging and SCILS Lab software are expensive. MALDIQuant was one of the first free statistical analysis packages for MALDI data, but it is only MALDI specific, not MALDI-MSI specific; interpretation is limited and does not contain the ability to view heat maps. Cardinal is a new R package that has been published in the last year that provides all the necessary components: basic data processing and multivariate statistical analysis of MSI data with the capability of visualizing heat maps of extracted m/z. A literature search of bioinformatic tools relevant to MALD-MSI revealed 28 tools for data conversion, visualization, basic spectral process, and statistical analysis and is
Table 2 – A summary of bioinformatic tools relevant to MALDI-MSI for use in data conversion, visualization, basic spectral process, and statistical analysis modified from Baker et al. 2016.
Abbreviations: F indicates a free software; P indicates a paid for software; B indicates a paid for MATLAB package; G indicates a downloadable graphical user interface software; R indicates a R Statistical package; M indicates a MATLAB package; W indicates an online web server
Software Source Purpose Ref
MMSIT (MALDI MS Imaging Tool) Novartis & Applied BiosystemsFG Acquisition 64
CreateTarget/AnalyzeThis! For Bruker TOF/TOFFG Acquisition and
Conversion
65
4000 Imaging Novartis & Applied BiosystemsFG Analysis 66
AMASS University of CaliforniaFG Analysis 67
Biomap Novartis & Applied BiosystemsFG Analysis 68
Datacube Explorer MS Imaging SocietyFG Analysis 69
EXIMS (EXploring Imaging Mass Spectrometry data)
University of MelbourneFM Analysis 70
msiQuant Uppsala UniversityFG Analysis 71
SpectViewer MS Imaging SocietyPG Analysis 72
Axima2Analyze National Institute of Physiological Sciences,
JapanFG
Converter 73
raw to imzML converter MS Imaging SocietyPG Converter 74
Cardinal Purdue UniversityFR Statistical Analysis 75
MALDIquant Institute for Medical Informatics, Statistics and
EpidemiologyFR
Statistical Analysis 76
Multimaging ImabiotechPG Statistical Analysis 77
omniSpect The Georgia Institute of TechnologyFM Statistical Analysis 78
OpenMSI Lawrence Berkeley National LaboratoryFW Statistical Analysis 79
MSight ExPASyFG Viewing 81
mMass Academy of Sciences of the Czech RepublicFG Viewing and Basic
Analysis
82
MSiReader North Carolina State UniversityPM Viewing and Basic
Analysis
83
FlexImaging BrukerPG Viewing and
Exporting
84
ImageQuest Thermo Fisher ScientificPG Viewing and
Exporting
85
MALDIVision Premier BiosoftPG Viewing and
Exporting
86
MITICS (MALDI Imaging Team Imaging Computing System)
University of LilleFG Viewing and
Exporting
87
MSI.R Unidad IrapuatoFR Viewing and
Exporting
88
TissueView Applied BiosystemsPG Viewing and
Exporting
89
MSImageView Novartis & Applied BiosystemsFG Viewing of
FlashQuant data
Bottom-Up Proteomics & Metabolomics in MALDI-MSI
Top-down proteomics is the study and identification of full, intact proteins.91,92 Top-down proteomics is used for the analysis of post-translational modifications , protein expression, cellular localization, and interactions.93 However, a large disadvantage of
top-down proteomics is that it is limited to analyzing simple mixtures of proteins and
struggles with large complex biological samples.91 There is a lack of protein fractionation methods developed for mass spectrometry.94 A solution to this problem is to analyze bottom-up proteomics or metabolomics.
Bottom-up proteomics is the study and identification of proteins after enzymatic digestion.21 Trypsin is commonly used to its high specificity of protein digestion at the C-terminal of arginine and lysine residues.95 On-tissue tryptic digestion of proteins for localized bottom-up proteomics was first introduced in 2007.96 Studies have shown it’s applicability in traumatized skeletal muscle80, forensic fingerprint analysis97, brain
ischemia98, and cancer99.
Metabolomics is the study of all metabolites in an organism.100 Metabolomics provides a unique advantage over proteomics in that it more closely resembles a phenotype of an organism.101 There have been various applications of MALDI-MSI for localization of
metabolites including: chronic respiratory diseases102, localizing leaf metabolites103, and cancer104,105.
Both bottom-up proteomics and metabolomics provide new insights into the biochemical mechanisms of an organism. Ideally, if sample size is not a concern, a combination of top-down proteomics, bottom-up proteomics, and metabolomics analysis would be performed.106
Challenges in the Field of MALDI-MSI
MALDI-MSI has a high potential for impact in the biological sciences due to its unique ability to spatially localize thousands of mass features from a single tissue section.
However, little standardization into sample preparation procedures has been
accomplished. Bottom-up proteomics poses its own challenges in adding additional steps to the sample preparation protocol to wash away interfering metabolites and to perform an on-tissue protein digestion. The analysis of metabolites can be useful to gain insights into the phenotype of an organism. Optimization of all sample preparation protocols is critical for ensuring a complete, in-depth analysis of as many compounds as possible within a tissue. Once acquired, statistical processing of data allows for unbiased confirmation of significant results.
Thesis Objectives
The three objectives for this thesis project are:
1. To develop a method for bottom-up proteomic MALDI-MSI analysis of prostate cancer.
a. This required optimization of experimental protocols for maximum detection of peptides. Experimental protocols included the optimization of metabolite washing, trypsin digestion incubation time, and the application of MCAEF.
2. To develop an unbiased bioinformatics workflow for simultaneous processing of MALDI-MSI datasets of multiple tissue sections using a thyroid hormone treated tadpole imaging dataset.
a. This required creating a R Statistics Software – Cardinal Package workflow. The processing workflow included spectral processing and multiple statistical analyses to find and confirm significant findings. Both spectral processing and statistical analysis required optimization of multiple parameters based on instrument type, data type, biomolecule analyzed, and experimental set-up.
b. This required testing to confirm statistical processing workflow was appropriate for simultaneous processing multiple datasets of control and thyroid hormone treated tadpoles for induced metamorphosis.
3. To develop a method for metabolomics MALDI-MSI analysis of FFPE colorectal liver metastasis samples.
a. This required optimization of sample preparation protocols to be able to detect metabolites in the tissues. This required tackling the issue of using xylene, that is needed for deparaffinization, which also removes most metabolites from the tissue samples.
Chapter 2: Method Optimization of Bottom-Up Proteomic
Analysis of Prostate Cancer by MALDI-MSI for Biomarker
Discovery
Introduction
It is predicted that 21,000 Canadian men will be diagnosed with prostate cancer in 2017. At 20.7%, prostate cancer has the highest incident rates of any type of cancer.107
Prostate cancer, along with all other cancers, emit unique proteins that can be used as biomarkers and detected in tissue and blood. Debate is ongoing whether or not the benefits of screening men for prostate cancer biomarkers are outweighed by potential risks of false positive or false negative assay results.108 Current screening uses an anti-body based method to detect the concentration of the biomarker prostate-specific antigen (PSA), and its derivatives, in serum. PSA was originally discovered and developed as a prostate cancer screening target in 1991.109 However, PSA tests are declining in use worldwide because of concerns of over cancer diagnosis due to false positive results. PSA tests cannot distinguish between prostate cancer, benign prostate hyperplasia or prostatitis.35 This results in unnecessary stress and exposure to invasive procedures. A
large meta-analysis of 53 individual studies reported over diagnosis rates ranging up to 67% for men of all ages worldwide.110 As a result, currently in Canada the PSA test is not recommended.111 There is a need for a biomarker and testing method that has greater sensitivity and specificity for detecting prostate cancer.
Previous studies from the Borchers laboratory on prostate cancer biomarkers included a top-down proteomics approach58, aimed at discovering significant intact proteins, and a metabolomics approach39, aimed at discovering significant metabolites. Figure 5 shows reconstructed ion heat maps of the top-down proteomic prostate cancer biomarkers discovered in the experiment.58 A total of 17 different proteins were shown to have
statistically significant differential expression in the cancerous region versus the non-cancerous region of the tissue. Three of these proteins were solely expressed in the cancerous region of the tissue (S100-A9, S100-A10, and S100-A12). Figure 6 shows representative reconstructed ion heat maps of metabolites shown to be of interest in this experiment.39 Over 600 metabolomic features were shown to have differential expression
Figure 5 - Reconstructed ion heat maps of proteins shown to have differential expression in the cancerous and non-cancerous regions of a human prostate cancer tissue. Proteins were detected with sinapinic acid as the matrix in positive mode58. Image also adapted from Baker et al. 2016.36 Images are reprinted with permission.
in the cancerous region when compared to the non-cancerous region. Two-hundred fifty metabolites were found only in the cancerous region, 217 metabolites were found only in the non-cancerous region, and 152 metabolites were shown to be significantly changed (p < 0.05, t-test) between the two regions.
Figure 6 – Reconstructed ion heat maps of metabolites (phospholipids and neutral lipids) shown to be of interest in a human prostate cancer tissues. Metabolites were detected with quercetin as the matrix in both positive and negative ionization mode39. Images are reprinted with permission.
These two experimental approaches left room for a third approach to be performed on the human prostate cancer tissues. The next logical step was to perform a bottom-up proteomics experiment in which the proteins from cancerous and non-cancerous tissues are cleaved with trypsin to create tryptic peptides for detection. Trypsin is a protease that specifically cleaves proteins at the carboxyl end of amino acids lysine and arginine. Trypsin is very useful for two main reasons: it creates mass fragments that are in a more favourable mass range for easier detection by mass spectrometers and are basic due to
being cleaved at the carboxyl side, creating an easily ionizable location on the peptide. Differential occurrence could enable identification of potential biomarkers.
Bottom-up on-tissue MALDI-MSI had not yet been performed in our laboratory, so no protocol existed. A literature review was performed to determine if standard protocols existed. A summary of this literature search can be found in Supplementary Table 1. Standard method preparation steps include: cutting the tissue, mounting the tissue onto a MALDI compatible surface, drying the tissue, washing the tissue to remove
ion-suppressing metabolites, trypsin deposition, trypsin incubation, and matrix deposition.
This data chapter focuses on the development and results of a bottom-up proteomic MALDI-MSI experiment on a human prostate cancer tissue. Optimization of sample preparation parameters found the following results: an additional wash step of 50mM sodium bicarbonate after the standardized steps (70% ethanol, 70% ethanol, 100% ethanol) was favorable due to increasing the pH of the tissue which is optimal for tryptic digestion; a trypsin incubation time of 18 hours to allow complete trypsin digestion; and to include the MCAEF technique with a CHCA matrix deposition for increased spectra quality. A total of 245 peptides belonging to 86 unique proteins were identified using accurate mass matching.
Methods
Samples
The prostate cancer sample used was obtained from BioServe BioTechnologies (Beltsville, MD, USA). The tissue was from a 62 ± 2-year-old male patient during prostate cancer surgical removal, with the patient’s informed consent. The tissue was diagnosed at prostate cancer stage II. The tissue was stored at -80 ˚C upon receipt and for the duration of the experiment. The use of anonymized human samples was approved by the Ethics Committee of the University of Victoria.
Reagents
Ammonium bicarbonate (NH4HCO3), sodium hydroxide (NaOH), dithiothreitol (DTT),
iodoacetamide (IAA), acetic acid, hematoxylin & eosin (H&E), high performance liquid chromatography (HPLC)-grade ethanol, HPLC-grade methanol,
liquid-chromatography/mass spectrometry (LC/MS)-grade acetonitrile (ACN), chloroform, xylene, water, trifluoroacetic acid (TFA), and formic acid (FA), quercetin, and CHCA were purchased from Sigma-Aldrich (St. Louis, MO, USA). Sequencing Grade Modified Trypsin was purchased from Promega GmbH (Product number: V5111, Mannheim, Germany). ESI tuning mix was purchased from Agilent (Agilent Technologies, Santa Clara, CA, USA).
Tissue Sectioning
The frozen prostate specimen was sectioned at -24 ˚C in a Microm HM500 cryostat (Waldorf, Germany). Serial sections of 14-µm thickness were thaw-mounted onto ITO-coated glass microscope slides obtained from Bruker Daltonics (Bremen, Germany).
Tissue Washing
As per the literature review presented in the introduction, there were many different wash steps used in various experimental protocols. A “basic wash” is a standardized wash protocol that consists of two sequential washes in 70% ethanol and one wash in 100% (or 95%) ethanol. All wash steps were performed for 30 seconds, unless noted. All washes were performed in 50 mL Eppendorf tubes. All solutions were prepared once for the duration of the method optimization experiment. After washing the tissue was complete, the slide was placed under vacuum for 10 minutes to ensure complete evaporation of the solvents. The following five washes were tested to determine which wash protocol was optimal:
1. basic wash
2. basic wash + 90:9:1 ethanol/acetic acid/water 3. basic wash + 50 mM NH4HCO3
4. Carnoy’s wash 1. 70% ethanol 2. 100% ethanol
3. 6:3:1 ethanol/chloroform/acetic acid (2 minutes) 4. 100% ethanol
5. water
6. 100% ethanol 5. no wash
Trypsin Digestion
A 20µg pre-weighed amount of trypsin was used for each slide. As per the product information instructions from Promega, 40-µL of 50 mM acetic acid was added to the Eppendorf tube and vortexed for 2 minutes. If necessary, this solution was stored at -4 ˚C until needed. As per the ImagePrep Instruction Manual, once ready for trypsin deposition, 200-µL 50 mM NH4HCO3 was added to the Eppendorf tube. This solution was vortexed
for another 2 minutes. Two-hundred twenty µL of the solution was transferred to the ImagePrep electronic matrix sprayer (Bruker, Bremen, Germany). The ImagePrep trypsin digestion deposition method was used.
Once trypsin deposition was complete, the slide was immediately transferred to a humidity chamber for the trypsin incubation. It was essential to avoid exposure to air to maintain trypsin activity. The humidity chamber apparatus was a heated water bath with a raised platform in the centre for the slide to sit on and a plastic cover over top of the bath. A small cover was placed on top of the platform to ensure that the slide was not
contaminated with falling water drops from the large plastic cover. Once the incubation was complete, slides were dried under vacuum for 20 minutes. The following times were tested for trypsin digestion times:
2. overnight (16-18 hours)
Matrix Coating
CHCA was prepared at a concentration of 7 mg/mL in 50% ACN and 0.2% TFA. Typically, a total of 10 mL was made at a time. Slides and tissue sections were coated with the CHCA matrix with the ImagePrep electronic matrix spray generator. The matrix coating was performed with the standard ImagePrep CHCA method.
During matrix coating, the previously described MCAEF technique was applied. A uniform electric field at an intensity of +600 V/m was applied to the tissue in positive-ion mode. An experiment was performed without MCAEF to ensure bottom-up proteomic mass spectra were also enhanced by the technique. This was done by performing a typical matrix coating with the standard ImagePrep instrument that did not have applied voltages or secondary conducive glass slide.
MALDI-MSI
All mass spectra were acquired on an Apex-Qe 12-Tesla hybrid quadrupole-FTICR mass spectrometer ((Bruker Daltonics, Billerica, MA, USA). The instrument was equipped with an Apollo dual-mode ESI/MALDI ion source. The instrument’s laser source was a 355-nm solid-state Smartbeam Nd:YAG ultraviolet laser (Azura Laser AG, Berlin, Germany).
A calibration solution was prepared by adding 10 µL of Agilent “ESI tuning mix” solution with 10 mL of 75% ACN with 0.1% formic acid. The calibration was directly infused into the ESI side of the ion source for instrument calibration. The instrument was calibrated before every experiment performed.
All mass spectra were acquired over the mass range of 200-2200 m/z with a data acquisition size of 512 kilobytes per second. Profiling data was acquired by accumulating twenty scans at 100 laser shots per scan. Profiling spectra were acquired with
ApexControl (Bruker, Bremen, Germany). Imaging data was acquired with one scan at 100 laser shots per scan with the minimum raster step size of 200-µm. Imaging spectra were acquired with Apex Control, Hystar Control, and FlexImaging 2.1 (Bruker, Bremen, Germany). Teaching points were generated with a Tipex Wite-Out pen to enable the instrument to have an accurate slide position for spectra acquisition. Profiling data was acquired for each of the different method steps during the method optimization. Profiling data acquired was recorded to be from the cancerous or non-cancerous regions of the tissue. Imaging data was acquired for the final optimized method protocol.
Data Analysis
Profiling mass spectra were processed using DataAnalysis 4.0 (Bruker, Bremen, Germany). Batch internal mass calibration, peak de-isotoping, and monoisotopic peak picking were performed using a customized VBA script within DataAnalysis. Peak alignment was performed with a previously written custom program with LabView
development suite with an allowable mass error of 3 ppm. The custom VBA
DataAnalysis script and customized program with LabView are described elsewhere.112
Imaging mass spectra were viewed with FlexImaging. Using this software, reconstructed ion heat maps were created. Images were exported as JPEG files for viewing. Principle component analysis (PCA) and partial least squares-discriminant analysis (PLS-DA) were performed with MetaboAnalyst 3.0
(http://www.metaboanalyst.ca).113
Data Processing Statistical Analysis
Peak lists (m/z and intensity) were exported from multiple replicates and profiling spots from each experiment. Peak lists were exported into CSV format and opened using Microsoft Excel. The total number of spectra included in the bar graphs represents the n-value. Average peak counts were totaled after removing all background signals and peaks that did not meet a signal to noise minimum. Background signal was removed by finding peaks that occurred in 80% of the spectra. Peaks with a signal to noise ratio less than 5 were removed. Peak intensities were averaged from the peaks remaining after the
background signal was removed and peaks with a signal to noise ratio of less than 5. Peak lists were imported into R Statistical114 using R Studio115. The mean intensities and counts were calculated along with corresponding standard error of the mean and plotted on a standard bar graph. Statistical analysis included performing a Kruskal-Wallis Rank Sum analysis of variance and a Whitney U post-hoc test. p-values from the Mann-Witney U test were used for determining significance.
H&E Staining
In-house H&E staining was performed. Each solution was in a glass Coplin jar. The same solvents were used for the duration of the experiment as multiple slides can be stained from the same set of solutions. The following protocol was used:
1. 95% ethanol for 30 seconds 2. 70% ethanol for 30 seconds 3. water for 30 seconds 4. hematoxylin for 2 minutes 5. water for 30 seconds 6. 70% ethanol for 30 seconds 7. 95% ethanol for 30 seconds 8. eosin for 1 minute
9. 95% ethanol for 30 seconds 10. 100% ethanol for 30 seconds 11. xylene for 2 minutes
Protein Extraction
Twenty-five mg of prostate tissue was used for protein extraction based on previously described protocols.116,117 The human prostate sample was homogenized in 200 µL of water with a Retsch MM400 mixer mill (Haan, Germany) with two 5-mm stainless steel balls for 2 x 1 minute at 30 Hz. Next, 800 µL of 1:3 chloroform/methanol solution was added. This was followed by another homogenization step for 30 seconds at 30 Hz. The stainless-steel balls were removed and then the tube was centrifuged at 16000g for 20
minutes at 4 ˚C (Eppendorf 5425 R, Eppendorf, Mississauga, Canada). The lipid-containing supernatant was discarded. The protein pellet was resuspended in 600 µL of water containing 3% formic acid. The tube was then vortexed for 3 times for 1 minute each to fully re-suspend the protein. The tube was again centrifuged at 16000g for 15 minutes at 4 ˚C. The supernatant was collected using 200 µL gel loading pipette tips and divided equally between two tubes. Protein solutions were dried using a Speed Vac (Savant SPD1010 Thermo Electron Corporation, Waltham, MA, USA). Tubes were stored at -80 ˚C until used.
Protein Digestion
One protein pellet was resuspended in 300 µL of 25 mM DTT/25 mM NH4HCO3 and
vortexed until dissolved to break the disulfide bonds. The protein solution was then allowed to incubate at 56 ˚C in a Thermomixer (Eppendorf, Hamburg, Germany) at 750 rpm for 50 minutes. Then the alkylation was performed by adding 300 µL of 100 mM IAA/25 mM NH4HCO3 and letting sit at room temperature in the dark under tin foil for
45 minutes. To quench the reaction, 15 µL of 1 M DTT/25 mM NH4HCO3 was added.
Next, 200 µL of 50 ng/µL trypsin in 25 mM NH4HCO3 was added. The digestion was
allowed to incubate over night at 37 ˚C. Once incubation was complete, the reaction was quenched with 800 µL of 0.2% TFA. The solution was then vortexed for 2 minutes. Next, a Solid Phase Extraction (SPE) was performed with Oasis HLB SPE extraction cartridges (200mg/3mL column, Waters Inc. Milford, MA, USA). The column was first conditioned with 1 mL of methanol then 1 mL of 0.1% TFA. Then 1 mL of protein solution was loaded onto the column. The column was then washed with 3 times 1 mL 0.1% TFA. The
protein was eluted with 3 time 600 µL 75% ACN in 0.1 % TFA. The eluates were pooled and dried with a speed vac. The dried protein was then reconstituted in 100 µL of 2% ACN in 0.1% TFA and then a 1 in 100 dilution in 2% ACN in 0.1% TFA to achieve no more than 1 µg on the column at a time with an 8 µL injection volume.
LC-MS/MS
Liquid chromatography tandem mass spectrometry (LC-MS/MS) was performed by loading 8 µL onto a Magic C18-AQ trapping column (100µm I.D., 2 cm length, 5 µm, and 100 Å; Michrom BioResources Inc., Auburn, CA) and separated on an in-house packed Magic C18-AQ capillary column (75µm I.D.×15cm, 5µm, 100Å) at a flow rate of 300nL/min using a Thermo Scientific EASY-nLC II liquid chromatograph. The mobile phase was 2% ACN in 0.1% formic acid (Solvent A) and 90% ACN in 0.1% formic acid (Solvent B). The flow rate was set to 300 nL/minute with one of the following elution gradients: 5% B to 45% B in 45 minutes, 45% B to 80% B in 2 minutes, and 80% B to 100% B in 2 minutes for a total run time of 49 minutes or 5% B to 40% B in 100 minutes, 40% B to 80% B in 5 minutes, 80% B to 100% B in 2 minutes for a total run time of 120 minutes. A built-in equilibration step was performed between each set of injections.
The chromatographic system was coupled to a Orbitrap Fusion (ThermoFisher, San Jose, CA, USA) which has a nano-flow ESI source. A user defined lock mass of
445.12002 m/z (a ubiquitous siloxane contaminant) was used for calibration. The ESI ion source was set to 2550 V with an ion transfer temperature of 275 ˚C. For full scans with no fragmentation, the Orbitrap detector was set to a resolution of 120000 with a normal mass range of 400-2000 m/z and no quadrupole isolation. A dynamic exclusion was
performed where ions were excluded after being detected 2 times if occurring within 5 seconds. Ions were excluded from the data for 10 seconds. Data was acquired in positive mode with profile data recorded. The charge states of the ions detected were limited from 2-5 with undetermined charge states not included. The intensity threshold was set to 5.0e4. Decisions for the acquired mass spectra included top speed data dependent mode with the precursor priority set to most intense. For MS/MS both collision induced dissociation (CID) and higher-energy collisional dissociation (HCD) were used with the following settings: collision energy of 35%, with an activation of 0.25 and multi-stage activation off, using the ion trap as the detector with a rapid rate, the automatic gain collector set to 2.0e3 with 1 microscan set to a maximum injection time of 300 ms, and collecting data in centroid mode. Exclusion lists were made by opening the acquired data in Proteome Discoverer 1.4.0.228 software (Thermo Scientific, Bremen, Germany) and creating a list of the high confidence ions. The following methods were run:
1. CID with 49-minute gradient 2. CID with 120-minute gradient
3. CID with 49-minute gradient and exclusion list 4. HCD with 49-minute gradient
5. HCD with 120-minute gradient
6. HCD with 49-minute gradient and exclusion list
LC-MS/MS Data Analysis
Raw data files were analyzed with Proteome Discoverer to generate peak lists for database searching. Protein identification was carried out with an in-house Mascot 2.2
server searching Uniprot_Human (all entries) with the following settings: allowable missed trypsin cleavages of 2, a maximum fragment tolerance of 0.6 Da, a maximum precursor tolerance of 10 ppm, a static modification of carbamidomethylation (C), and dynamic modifications of deamidation (N, Q) and oxidation (M). Percolator was used to validate peptide assignments with the following settings: maximum delta Cn of 0.05, a strict false discovery rate of 0.01 and a relaxed false discovery rate of 0.05. All six acquired datasets were exported to excel together.
Accurate Mass Matching for MALDI-MSI Peptide ID and Protein Assignment
A general workflow for accurate mass matching can be found in Figure 7. Exported peak lists from MALDI-MSI data and exported peak lists from LC-MS/MS data were color coded and combined into a single excel column. After sorting all m/z values from highest to lowest, a ppm calculation was performed between all neighbouring MALDI-MSI and LC-MS/MS values. All m/z values that matched with under 5 ppm error were considered to be an accurate mass match for peptide identification and protein
Figure 7 – Workflow used for accurate mass matching between MALDI-MSI and LC-MS/MS data for peptide identification and protein assignment
Results & Discussion
Tissue Wash Optimization
After performing a literature review, it was apparent that different laboratories used various wash protocols to remove metabolite and other organic small molecules from the tissue that may cause ion suppression. The four most common washes were compared against a dataset with no wash step. The no wash step was performed to act as a baseline to determine the total amount of ion suppression caused by the background molecules. As seen in Figure 8, all washes except the basic wash significantly increased the average
Figure 8 – (a) Average peak intensity and (b) average peak count of the MALDI-MSI wash optimization experiment on prostate cancer tissue. Detail for each wash can be found in the Methods section. Not all significant comparisons shown on graph. Detailed p-values can be found in Supplementary Table 2.* p < 0.05, ** p < 0.01, and *** p < 0.001
peak intensity (p-values <2e-16, <2e-16, and 1.9e-5, respectively). All p-values
calculated by a Mann Whitney U test can be found in Supplementary Table 2. Of note is the significant increase of the peak intensity of the basic wash + 50 mM sodium
bicarbonate in comparison to the no wash experiment and the basic wash + 50 mM acetic acid wash. It is predicted that the increased peak intensity of the sodium bicarbonate wash is due to decreased ion suppression due to background signals from lipids and other metabolites. The increased peak intensity of the sodium bicarbonate wash compared to the acetic acid wash aided in the final decision of using a sodium bicarbonate wash (p-value 0.017). A significantly higher peak count for the no wash experiment compared to the basic wash and Carnoy’s wash (p-value 0.0093 and 0.0015, respectively) amplifies the prediction of having background and metabolite signals in the spectra. An additional reasoning behind choosing the sodium bicarbonate wash was because the step immediate after the wash step is trypsin digestion, which performs optimally at a slightly basic pH, which is aided by sodium bicarbonate rather than the acetic acid or Carnoy’s wash. The method preparation and instrument parameters were optimized for peptides, meaning peak intensities may not be strong for metabolites. However, they will still be detected and can cause not only an artificially high peak count, but cause ion suppression for peptide signals. Ideally the sodium bicarbonate wash would have shown the highest peak intensity and highest peak count. Unfortunately, the acetic acid wash showed a
significantly higher peak count than the bicarbonate wash (p-value 0.0034). Due to the trypsin digestion potentially being affected by the pH, the sodium bicarbonate wash was given preference. It was confirmed that the sodium bicarbonate wash showed
significantly higher peak counts in comparison to only using a basic wash (p-value 0.0034).
Trypsin Digestion Incubation Time Optimization
The literature review also revealed an inconsistency in trypsin digestion incubation time. Two hours was more common for formalin-fixed and paraffin embedded (FFPE) tissue trypsin digestion, however it was used occasionally for fresh-frozen tissue.
Overnight digestion (18 hours) was more common for fresh-frozen tissue. Comparison of peak intensities and peak counts gave for differing results (Figure 9). The overnight
Figure 9 – (a) Average peak intensity and (b) average peak count of the MALDI-MSI trypsin digestion incubation time experiment on prostate cancer tissue. Details for each incubation time can be found in the Methods section. Detailed p-values can be found in Supplementary Table 2.* p < 0.05, ** p < 0.01, and *** p < 0.001
